As AI becomes more capable, businesses may find themselves asking a familiar question: How much of a process should they change, and how much should they keep under direct human control? That question becomes especially important when AI connects to an enterprise resource planning (ERP) system. A response about a purchase order may seem simple, but changing that order could affect a supplier commitment, production schedule, and financial records.
Over the next couple weeks, we will look at when AI should be allowed to act, what happens when customer-facing recommendations reach ERP operations, and how organizations can incorporate AI into existing processes. Throughout our exploration, we will dive into real-world cases and consider how their findings could shape your organization’s future.
When Should an AI Agent Be Allowed to Act in ERP?
Suppose a purchasing employee asks an AI assistant why the price on a purchase order changed. The assistant might review the order, supplier correspondence, and contract terms, then explain what it found. From there, it could draft a correction for the employee to review. But should it be allowed to submit the change or approve it?
Each step gives AI a different level of responsibility. As a result, organizations need to decide where the approval boundary belongs for each process:
-
- Answer: Explain its status to an authorized employee.
- Draft: Prepare a proposed change for review.
- Act: Submit or approve a change that affects the ERP record.
A study on conversational AI in ERP illustrates the information side of this question. In one financial business insights scenario, a project controller is asked to investigate why a project’s recognized margin fell short of expectations. The conversational tool helps analyze ERP financial information. The researchers also describe a way to display the search results used for an answer, giving the employee an opportunity to check the underlying material.
On the other hand, an agentic AI study shows what happens when the system goes beyond answering. Its billing dispute example begins with a customer questioning an unexpected charge. The agent creates a case, compares the bill with contract terms, identifies an overcharge, and recommends a credit note. In the proposed workflow, it could issue a credit automatically if the amount falls below a predefined approval threshold, while recording its actions in the case notes.
Where Does CRM AI Meet ERP Reality?
Once an organization sets limits on AI actions, leadership needs to ask whether its recommendations reflect the full business picture.
Consider an illustrative example: A manufacturing company’s sales tool predicts increased demand and suggests that a customer can receive an order earlier than planned. The recommendation may sound reasonable from a sales perspective.
- But is the required inventory available?
- Does the production schedule have room?
- Can the fulfillment team meet the new date without delaying another customer?
The request then continues into pricing, invoicing, and finance. If the promised terms do not match the ERP record, a quick answer to the customer can turn into a delivery problem or billing dispute. Because of this, the connection between customer relationship management (CRM) and ERP deserves attention. AI may recognize an opportunity in one part of the organization, while missing a constraint in another.
Muhammad Faizan Hassan’s review of AI in ERP discusses applications across CRM, manufacturing, supply chain, and finance. It also proposes looking at organizational readiness, phased implementation, governance, and performance measurement together. In an upcoming article, we will follow that journey to examine where a forecast or customer request needs to be checked against operational data, and how leaders might measure outcomes such as reliable delivery commitments and fewer invoice exceptions.
How Can Organizations Incorporate AI into ERP?
After identifying a business need, where should an organization begin? Leadership may want to start with a platform, or a broad goal such as, “Automate purchasing.” However, that goal does not tell a team which information AI needs, what decisions it should support, or what it should be permitted to change.
The framework in A model for incorporating AI into ERP software offers a way to break down the process. It looks separately at how data is acquired, how information is analyzed, how decisions are made, and how actions are executed. Applied to purchasing, a company might first make order and contract information easier to retrieve. It could then use AI to flag discrepancies and prepare proposed changes, while buyers continue to approve updates. The business can then assess the results before expanding AI’s authority. The study recommends assessing the current process and defining a target with business stakeholders. For instance, a relatively stable manufacturing workflow may benefit from a focused improvement, while a variable demand-planning process may call for more advanced prediction.
However, the question then becomes, “What outcome are we trying to improve, and which part of the process needs AI to achieve it?”
Testing should be part of that answer. The proposed Autonomous ERP Testing Framework Using Multi-Agent Artificial Intelligence describes specialized agents for finding processes, generating and running tests, analyzing defects, checking compliance, and assessing release readiness. Additionally, it provides an example of how AI might assist the testing effort. However, please note that business and technical teams still need to verify normal transactions, unusual cases, access permissions, and effects on connected processes before approving a change.
Do you understand where your organization is today, which outcomes need to improve, and where human judgment needs to remain? If so, you have already taken the first step to AI/ERP transformation! From there, AI and ERP integration becomes a set of decisions your teams can examine, test, and improve over time.
Stay tuned for upcoming articles that will dive deeper into the ERP/AI process! In the meantime, feel free to take a look at our previously published articles:
How to Integrate AI within Your ERP System
Leave A Comment